Not under the repository’s code licence. The pipeline is MIT; the transcript text is not. It is machine-generated from recordings of the Shell Game podcast episodes, and those words belong to the speakers and rights holders. Read CONTENT_LICENSE.md before republishing, redistributing, or building anything on it. Quoting with attribution to the original work is ordinary practice and unaffected.
Accurate enough to search, not accurate enough to quote without checking. Whisper misrenders names, technical terms, and crosstalk in particular. The summaries are a further lossy step on top of that. Where accuracy matters, go to the original recording.
Most likely the pipeline has not been run for that year, or the stage failed partway and its resume log records the episodes that did succeed. Re-running is safe — completed work is skipped. See Pipeline.
The resume log still lists it. Stages consult the log, not the filesystem. Delete the relevant .log entry or the whole log — see Pipeline.
Transcription. Whisper on a CPU can take longer than the episode’s own running time. Installing the CUDA build of PyTorch is the single biggest improvement available — see Installation.
Yes, though not without editing code. The Whisper and Ollama models are literals inside the stage scripts; Configuration lists exactly which line. Remember to clear the relevant resume log, or nothing will be regenerated.
No. Everything either runs locally or reads public feeds. See External Services.
Open an issue or contact the repository owner. CONTENT_LICENSE.md commits to honouring removal requests without argument or delay.